40 THE STUDY OF SPEECH CURVES. 



than from the reproducing; this is due to such facts as that the reproducers 

 are less well made than the recorders, that the reproducer point does not 

 have time to follow the details of the curve, that details are lost in multiply- 

 ing records, etc. When a record is heard to say distinctly " Minots Ledge 

 Lighthouse" we can feel sure that the curve traced from it contains the 

 waves for [mainats], etc. We can also be sure that it contains many 

 details more for these sounds and their glides than the ear can possibly 

 detect, and that the departure of the waves from absolute accuracy is 

 less than that of the sounds heard from the original sounds. 



With the accurately traced record before us we can undertake the 

 study of the curve by the unaided eye. For a phonetician accustomed 

 to interpret curves, the results of such a "qualitative analysis" are so 

 rich that they alone amply repay the cost and pains of tracing records. In 

 this respect the method stands in strong contrast to " mathematical analy- 

 sis." Here the expert gathers in a few hours a mass of facts of great 

 importance, whereas the full mathematical analysis of a single wave of 

 a vowel occupies days and furnishes only one minute fact; to be sure there 

 is only one way to get facts of the latter kind, and the labor can not be 

 avoided if certain problems are undertaken. 



The best way of teaching qualitative analysis is to carry through in 

 detail a few examples; I have selected for this purpose a plate from the 

 Depew record and another from the Cock Robin record. 



The Depew plate is number 6 of a series giving the curves of his " Speech 

 on Forefathers' Day." Line 94 shows the weak vibrations for [w] followed 

 by the stronger ones for [i] ending with the first of the weaker ones for 

 [S]. The [5] extends nearly to the middle of the line 95, the vibrations 

 becoming steadily weaker. Here the [au] begins; it continues through 

 line 96 until it glides into [t] in hue 97. In Une 97 a piece 205mm. long 

 of straight Une has been omitted. Let us examine the curve somewhat 

 in detail. 



To understand a vowel curve we note that the vibrations occur in 

 groups; for example, in the middle of hue 96, the vibrations fall into a 

 series of wave groups. Each group represents the result of one puff (or 

 vibration) from the glottis after acting on the set ofi vocal cavities. 

 Since each group corresponds to one puff from the glottis, the length of 

 the group can be used to give the period of the puff and consequently the 

 pitch of the voice. Shorter groups represent higher tones, longer groups 

 lower ones. Without making any measurements we can compare groups 

 by means of dividers. We can at once make two observations. In the 

 first place, the length of the groups is not constant; this means that the 

 voice changes its pitch at each instant. This holds good of all speech 



